A Petri-net Based Approach To Modeling And Quality-of-service Evaluation Of Service Composition

Web services is the self-contained and self-describing modular application procedure that can be called by interface.. As a new type of distributed component model, Web services is playing an increasingly important role on e-commerce, enterprise application integration and other fields, especially, since it can achieve the services'reuse and value-adding, Web service combination technology has become focus of academic and industry.Service combination is a kind of technology to reuse services by viewing services as components; it can select particular services from the existing services, and combine them into a new service to meet the needs of users. Currently, service combination protocols are tend to be unity, and mainly in the following three kinds of technique: the Business Process Execution Language(Business Process Execution Language for Web Services, BPEL4WS/BPEL) which combined by XLANG and WSFL (Web Services Flow Language); business processes Modeling Language (Business Process Modeling Language, BPML); W3C's Web service orchestration definition language (Web Services Choreography Definition Language, WS-CDL).A number of uncertainties caused by the services characteristics of distributed, heterogeneous, autonomous and dynamic. Firstly, services participated in business process are widely distributed on the Internet, possiblely, since the coordination , matching and interoperability problems among different organizations, different systems platforms involvs many fators, any problem of details any possible to undermine the stability of the whole; Meanwhile, the dynamic release, failure to remove,delete of individual service components, ,will lead to service quality attributes of individuals may change at any time, also led to the difficult prediction of the overall quality of the services portfolio; and the hardware and software conditions and network environment which the running of service combination relied on will also occur real-time change, resulting in the overall quality hard to maintain stability.Therefore reliability analysis of service combination become necessary for promoting the further development of WEB services technology. Both Academic and industry area urgently need metholds and tools of modeling, analysis, evaluation, validation and optimization their reliability According to the authority definition of research on software reliability, software reliability is defined as the degree to satisfy the functional and non functional requirements of software (that is, Quality-of-service). This study focused on the latter of the above two aspects, comprehensively using the formal methods, probabilistic analysis and stochastic processes, confidence interval analysis theory and technology, model, analyze and evaluate the service quality of services combination. The research is expected to propose certain significance analysis methods and theoretical results in this area, and has great practical significance on the improvement and development of WEB services technology.Specific work of this article include:(1) Design conversion method, realizing the complete conversion from Web Services Business Process Execution Language which based on the orchestration mechanism and WS-CDL that based on choreography mechanism to the meidiate model (stochastic PETRI net), and completely descript important details of services combination's process control, interoperability, process calls, message interaction, remove the dead end, exceptions / compensation / error handling, and so on. The above conversion rules not only can model the control flow when the service combination process being normaly executed, but also can describe the impact of unexpected conditions (error / event / compensation / termination processing activities) on the process of evolution. It overcome some shortcomings of the existing research, such as the research model is grained, and the conversion is not complete.(2)Based on the intermediate model of Stochastic Petri nets, conducting status analysis on the control flow evolution of Service combination, embedded in continuous time Markov chain , and obtain the intensity matrix corresponding to changes Then design a series of algorithms, and calculate in the change intensity matrix, including three service quality indicators : the Expected Process Normal Completion Time , the Process Normal Completion Probability, and the Expected Overhead of Normal Completion.(3) In order to test the accuracy of theoretical results and the correctness of the method based on stochastic Petri nets, we use WS-CDL + / ACTIVE BPEL execution engine to test the practice example of WS-CDL/WS-BPEL, and extracted data related to the three service quality indicators from the logging file .Samples by the above data, we use the method of confidence interval, results show that the three indicators all have corresponding 95% confidence interval. Compared the results of confidence interval to results obtained from method based on Petri net theory , we found that confidence interval coverage the results of theoretical analysis entirely, which verifies the accuracy of the results and correctness of the analysis method.The work involved in this paper,overcome the shortcomings of the research on WEB services combination theory in the area of service quality and quantitative analysis , it can provide automated analysis technology of model support and means of quantitative analysis for WEB Service reliability analysis, and is expected to make breakthrough in the weak WEB service area, has positive significance on establishing our service-oriented computing technology independent innovation system. In particular, its value including three aspects:(1) realizing a combination of the two mainstream services agreement (WS-CDL/WS-BPEL) of fine-grained model, completely characterized the important impact factors of service combination and execution and control flow evolution ;(2) realizing the conversion of the WS-CDL/WS-BPEL combination protocol to random Petri net ;(3) analized the most three important service quality indicators , and verified the correctness and accuracy of the results.